awangwang123/jianhao-travel-planner

出行路书工作流 skill：联网实查 + 多源交叉验证，产出可核验、能执行的旅行攻略。覆盖吃住行游拍避全维度，附美食情报卡、基准骨架与校验工具，支持一键部署在线版。

见好 · 旅行规划器

What it is – A Claude‑AI skill that turns a simple request like “I want to travel to X this weekend” into a complete, offline‑readable travel road‑book. The skill follows a fixed workflow: it gathers the user’s preferences, fetches up‑to‑date information from public sources, and produces two artefacts:

A single‑file HTML road‑book (full version that works without internet, plus a lightweight online version you can deploy yourself).
A “food intelligence card” that accumulates city‑specific dining data over time.

The output is a richly structured document with ten sections – overview, transport, daily itinerary, ticketing, tips, accommodation, budget, emergency, food list, and local specialties – each populated with real photos, hand‑drawn route maps, and verified price/availability data.

What you get out of the box

📄 One‑file HTML road‑book – looks good on desktop, mobile, or printed, and can be opened completely offline.
🌐 Optional one‑click online deployment – a slim HTML version you can host on your own server.
🔍 Verified information – key facts (prices, opening hours, transport options) are cross‑checked against live sources (e.g., Dianping, Gaode, Xiaohongshu) and include source citations.
🖼 Real‑world photos + hand‑drawn maps – no generic stock images; the skill pulls recent Wikimedia images or, if you prefer, Pexels or your own AI‑generated placeholders.
🍜 Structured food recommendations – per‑day, per‑meal lists with must‑try dishes and average cost.
✅ Built‑in checklist – tools/checklist.py runs 11 automated consistency checks before the AI finalises the document, ensuring nothing is missing.
🔒 No personal data leakage – the skill never stores sensitive user information.

How it works (high‑level flow)

Trigger – Speak a trigger phrase to Claude (e.g., “周末想去厦门玩”).
Preference collection – The skill reads the user profile template in SKILL.md (departure city, food preferences, travel style, etc.).
Live data fetch – Queries public APIs / web sources for weather, transport options, ticket prices, restaurant reviews, etc.
Document generation – Populates a Markdown source (md fact file) and then renders two HTML templates located in assets/.
Self‑check – Runs tools/checklist.py (and tools/desource.py for data sanitisation). If any of the 11 checks fail, Claude is prompted to fix the issue.
Delivery – You receive the single‑file HTML road‑book and the food intelligence card.

Quick installation (≈2 minutes)

Download the repository and unzip it. The top‑level folder must stay named jianhao‑travel‑planner.
Place the folder in Claude’s skill directory:
- Global: ~/.claude/skills/ (Windows: C:\Users\<User>\.claude\skills\).
- Or project‑local: <your‑project>/.claude/skills/.
Verify the path – you should see …/skills/jianhao-travel-planner/SKILL.md.
(Optional) Set a Pexels API key if you want to use Pexels images: add PEXELS_API_KEY=your_key to a .env file at the repo root.
Fill your profile – open SKILL.md, scroll to the “用户档案（填空模板）” section, and complete the placeholders (home city, dietary preferences, typical travel pace, etc.).
Talk to Claude – say a trigger phrase like “周末想去成都玩” and the skill will start the workflow.

Who might use it

Travel enthusiasts who want a polished, printable itinerary without manually stitching together information.
Content creators / travel bloggers needing a reproducible, source‑cited road‑book they can share with followers.
Small agencies or freelancers that want to automate itinerary generation while keeping full control over data and branding.

Current status

Actively maintained (latest release v1.7 on 2026‑10‑08).
Open‑source under the MIT license.
Community‑driven feature requests via GitHub Discussions (e.g., “city wish‑pool”).

TL;DR – 见好·旅行规划器 is a Claude‑AI skill that, with a single spoken request, produces a beautifully formatted, offline‑ready travel guide complete with verified data, real photos, and a food‑recommendation card, all without needing paid APIs.